Embodiments in accordance with the invention relate to optical sensors and image devices and, in particular, to a detector for detecting electromagnetic radiation and to a method for detecting electromagnetic radiation, and to a method for producing a detector.
Optical sensors and image devices are important objects in many fields of everyday life. In consumer applications, such as, for example, digital photography and mobile phone cameras, simple image devices have gone into industrial mass production. Nevertheless, there is a huge market potential for special image devices, like for example in the fields of monitoring or in industry, medicine, in the field of motor vehicles and in scientific applications.
Within these fields, there is a great diversity of application-related requirements which are still the object of active research, such as, for example, highly sensitive sensors for operating under poor light conditions, infrared cameras, 3D cameras providing distance information in the image, or high-speed cameras, all of which are combined with the requirement of cost efficiency and reliability.
In order to address the diversity of requirements for many different applications, very large versatility and flexibility are needed in both technologies and design topics, in particular when the image devices are exemplarily to be generated in CMOS (Complementary Metal Oxide Semiconductor) technologies, which allows in-pixel signal processing, x-y-pixel addressing possibilities, the “camera-on-chip” approach and low cost compared to other solid-body imaging technologies. For 3D type-of-flight measurements or applications for machine vision, typically large-area low-noise high-speed photodetectors are needed. In this type of application, the number of pixels of a sensor or the miniaturization thereof is not the main focus.
The development of pixel configurations using active photo control electrodes (photo gates, PGs), produced in CMOS technology in a “camera-on-chip” approach, started in the early 1990s based on the known and fully-developed CCD (charge couple device) technology. Although these were not new, using same allowed for the development of a wide range of different CMOS image sensor configurations over the years, as was the case before with the p-n junction-based photodiodes. The PG-active pixels based on CMOS technology offer several advantages compared to conventional pixels based on p-n junction photodiodes. The main advantage is improved noise suppression caused by the non-destructive readout thereof. The kTC noise at the pixel output here is not dependent on the capacity of the photodetector. It is defined by the much smaller capacity of the readout node, the so-called floating diffusion (FD) which is charged-coupled to the PG by a MOS capacity-based control electrode (transfer gate, TG). When there are a separate photo-active region and a readout node region, this allows integrated charge readout of the floating diffusion (FD), whereas additionally charge-to-voltage conversion and amplification may be performed. Nevertheless, using a polysilicon layer over the photo-active region reduces its quantum efficiency, in particular in the blue and ultra-violet ranges of the spectrum.
On the other hand, the pixel filling factor decreases with an increasing complexity in the pixel structure. Thus, a compromise has to be found. As an example of application, the photodetectors in a time-of-flight (ToF) 3D imaging application are to be highly sensitive, in particular in the near infrared range (NIR) of the spectrum, the signal-to-noise ratio (SNR) thereof is to be very high (the pixel noise is to be minimized), and the response speed of a pixel also is to be high, in particular when measurements are performed using laser pulses, the pulse periods exemplarily being TLaser≅100 ns.
Two types of currents are to be taken into consideration when analyzing the speed performance of the PG-based pixel structure: drift and diffusion currents which are induced by a charge transfer from the photo control electrode or gate (PG) to the floating diffusion (FD) via the transfer control electrode or gate (TG). An electrical drift field under the photo gate and the transfer gate allow collected photo-generated charge carriers to be transferred, however, only when there is an electrostatic potential gradient which generates a drift field. For longer photo gates (PGs), the electrostatic potential under the photo gate remains constant, which means that the minority carriers which are collected under the photo gate, can only be transported to the floating diffusion (FD) by means of thermal diffusion, which makes transport very slow. The transfer and readout times of roughly 20 μs are normally achieved in this type of photodetectors using short integration times of 5 μs.
FIG. 7 shows a schematic illustration of a known detector 700 comprising a photo control electrode gate (PG) 710, a transfer control electrode or gate (TG) 720 and a floating diffusion (FD) 730. Additionally, a readout circuit and a reset control electrode (RST) are indicated. Correspondingly, FIG. 8 shows a schematic illustration of an oscilloscope graph 800 of signals of the detector of FIG. 7. The oscilloscope graph 800 shows the reset, PG and TG signals and the voltages at the source follower output under illuminated (λ=700 nm, E=1.69 e−3 W/m2) and dark conditions.
On the other hand, using p-n junction-based photodiodes generates a relatively large amount of equivalent noise charge (ENC) in the photodetector due to its relatively high capacity compared to the previously defined floating diffusion (FD), and the collected charge in these types of applications is held at the silicon surface where an additional noise portion is added to the signal charge. Furthermore, no suitable CDS (correlated double sampling) techniques can be applied in this kind of structure since the charge collecting and readout regions are united here. This kind of application results in unacceptable signal-to-noise ratio numbers.
The so-called “buried” photodiodes and “buried” photo control electrodes or gates are, as has been proven, a good solution for reducing the amount of noise in a photodetector, since the electrostatic potential maximum in these photodetectors is pushed away from the silicon surface, however, they still exhibit problems as far as response speed and overall well capacity are concerned.